Every digital agency in the UK now offers generative engine optimisation. Most of them are measuring the same things they always measured and calling it something new.
That is the problem you face when you sit in a pitch meeting. Two agencies describe GEO. One talks about structured data and schema. The other talks about brand mentions and third-party coverage. You leave the meeting wondering whether these are the same service, completely different workstreams, or whether either agency genuinely knows what they are doing.
You are not alone in that confusion. And you are not wrong to be sceptical.
What this guide covers:
-
What a generative engine optimisation agency actually does, and what a real engagement looks like
-
What the first 90 days of a credible programme should contain
-
How to measure GEO properly, and why most agencies miss this entirely
-
The five capabilities a GEO agency must have
-
Six questions that separate genuine expertise from rebranded SEO
-
When GEO makes sense, and when it does not
What a generative engine optimisation agency actually does
Generative engine optimisation is the practice of making your brand appear, by name, in the answers that AI systems generate. When someone asks ChatGPT which accountant to use in Birmingham, or asks Perplexity which agency handles AI search for professional services firms, or reads a Google AI Overview comparing software options, GEO is the discipline that determines whether your brand is cited, recommended, or quoted in that response.
This is meaningfully different from traditional SEO, which is about placing your pages in a list of links. GEO is about whether an AI model, synthesising information from across the web, treats your brand as a credible answer to a relevant question.
A genuine AI SEO agency running a full GEO programme works across four workstreams.
Entity optimisation. AI systems understand the world through entities: named people, organisations, locations, products, and concepts, and the relationships between them. Your brand is an entity. GEO requires ensuring that AI systems have consistent, accurate, and credible information about who you are, what you do, and why you are worth recommending. This means structured data implementation, consistent naming and location data across the web, schema markup that machine-readable systems can parse, and a coherent presence in the sources that AI systems draw from.
Content restructuring for AI readability. AI models extract answers from content that is structured for extraction, not just for human reading. Answer-first paragraphs, FAQ sections with clear questions and concise standalone answers, definitional content that opens with a precise explanation of the topic: these are the formats AI models quote from. Standard blog writing, crafted for scroll and skim, does not translate well into AI citation. Restructuring existing content to perform in this context is one of the most undervalued workstreams on the SERP right now.
Citation and authority building. The sources AI models trust most are established publications, industry bodies, and credible third-party websites that have independently referenced your brand. Digital PR, link acquisition, and trade press coverage all contribute to building these signals. Without a working digital PR capability, an agency cannot build the authority signals that drive AI recommendation. This is where many agencies claiming GEO expertise have a visible gap.
AI citation monitoring. Tracking how often your brand appears across a defined set of target prompts, in ChatGPT, Perplexity, Google AI Overviews, and other AI engines, over time. This is both the starting point of any programme and the ongoing measurement layer. An agency that cannot show you this data is not running a GEO programme. They are running content marketing and calling it something new.
What GEO is not: a one-time structural fix, a social media strategy, basic keyword targeting given a new name, or a guaranteed result in 30 days. Any agency framing it as a quick win is either confused or hoping you are.
What the first 90 days of a GEO engagement should look like
The first thing a credible GEO agency should do is establish a baseline. Before any optimisation work begins, they need to know how your brand currently appears across the AI engines they are targeting. Without that starting point, there is no way to prove that anything they did made a difference.
This is not a nice-to-have. It is the foundation of the whole engagement.
The baseline involves running 20 to 30 target queries through ChatGPT, Perplexity, and Google AI Overviews, recording the outputs in a consistent format, and mapping how often your brand appears, in what context, and alongside which competitors. Done properly, this takes time. An agency that skips it is either taking your money for work they cannot prove, or assuming you will not ask.
A simple test in the pitch meeting: ask the agency to run a basic citation check for your brand or sector right now, in the room. A capable team can open a laptop, type a relevant prompt into ChatGPT or Perplexity, and walk you through the output. If they tell you this is "part of the onboarding process," that is worth noting.
Month 2: Structural work. Once the baseline exists, the second phase is foundational. Entity mapping ensures your brand, services, key people, and locations are described consistently across the web. Schema implementation covers FAQPage, Organisation, and HowTo markup, deployed correctly. Content restructuring means rewriting key pages and articles to answer-first format, building FAQ sections with H3 question structure, and adding definitional content that AI models can extract and quote. None of this produces immediate citation lift. But without it, citation building later has nothing to anchor to.
Month 3: Citation building and the first measurement review. The third phase combines active digital PR outreach and link acquisition with the first comparison against the baseline. The relevant question at this review is not "did organic traffic go up?" but "are we being cited more often, in more relevant prompts, than we were at month one?" That is a different question. It requires different data.
Realistic expectations: measurable improvement in AI citation across target prompts typically appears at 60 to 120 days from programme start. The specific timeframe depends on your current domain authority, your sector's competitive intensity in AI outputs, and how much structural work was required in month two. Anyone promising visible results in 30 days is working with a different definition of "results" than the one worth measuring.
That timeline is longer than most agencies will tell you in a pitch. It is also more honest.
How to measure whether GEO is actually working
The most common failure mode in GEO measurement is using the wrong metrics. Organic traffic rankings tell you how your pages perform in traditional search. A proprietary "AI visibility score" from an agency's own dashboard tells you whatever the agency has decided to put in it. Neither is the core GEO metric.
Three things actually matter.
Branded mention rate. How often does your brand appear when a defined set of target prompts is run through ChatGPT, Perplexity, and Google AI Overviews? This is tracked by running the same 20 to 30 prompts each month and recording how often your brand appears, in what context, and in what position within the response. Averi's published measurement framework suggests a realistic target citation rate of around 10 to 20 percent across a defined prompt set, though this varies significantly by sector and competitive density. What matters more than the absolute figure is the direction of travel and the rate of change against a documented baseline.
AI referral traffic in GA4. Sessions arriving from perplexity.ai, chat.openai.com, bing.com/chat, and similar AI platforms. GA4 does not separate this automatically into a distinct channel, so you or your agency needs to configure a custom channel grouping to capture AI referral traffic separately from standard organic. Without this setup, the traffic is invisible or buried in the organic bucket. If your agency has not done this configuration, it is a simple first fix, but its absence tells you something about their measurement culture.
Conversion rate from AI sessions. Early data suggests that AI search traffic tends to convert at higher rates than traditional organic, because the user has already had their question answered and is arriving to act on it. Tracking conversion rate specifically for AI referral traffic gives you the clearest ROI picture and the strongest internal case when budget decisions come up.
Measurement cadence: the consensus among measurement specialists is that a two-week cycle is more responsive than a monthly one. AI outputs change quickly and a monthly snapshot misses the signal. For clients who need scale and consistency, specialist platforms including Averi and Conductor offer automated citation monitoring dashboards. For smaller programmes, manual prompt testing is entirely viable and costs nothing beyond the time to do it consistently.
If an agency's monthly report shows organic rankings, keyword positions, and a traffic graph with a GEO label on it, they are not measuring GEO. According to analysis from Search Engine Land, traditional search metrics and AI citation metrics are increasingly divergent as zero-click AI responses become the norm for high-intent queries. That divergence is exactly why measuring the right thing matters.
The capabilities a capable GEO agency must have
GEO is a multidisciplinary service. An agency strong in one area but weak in the others cannot deliver a complete programme. Here is what to look for, and what to probe in the pitch.
Technical SEO. GEO is built on top of SEO infrastructure. Schema implementation, crawlability, structured data, site architecture: these are prerequisites, not optional extras. An agency that cannot handle technical SEO correctly cannot implement the structured data that helps AI systems understand your brand as an entity. Ask about their technical SEO capability specifically, and ask who on their team does this work. If the answer is vague, that is meaningful information. A solid SEO agency foundation is what GEO builds on. Without it, you are optimising for AI visibility on unstable ground.
Content architecture. Writing for AI extraction is a distinct skill from standard content marketing. Answer-first structure, FAQ sections with H3 questions and concise self-contained answers, topical authority built through cluster content: these require both strategic planning and editorial craft. Ask to see examples of content they have restructured specifically for AI readability, not just content they have published.
Digital PR. The citations AI models trust most come from credible third-party sources: trade press, industry publications, high-authority websites that have independently referenced your brand. Without a digital PR capability or a working relationship with a digital PR partner, there is no clear path to building these authority signals. Ask specifically how they build third-party citations. If the answer is "we publish content on your site," they are missing a workstream entirely.
Measurement infrastructure. Can they show you actual AI citation monitoring output from a current client? Not a traffic graph. Not a keyword ranking table. The actual prompt-and-response data showing how a client's brand appeared in ChatGPT or Perplexity over a defined period. If they cannot produce this in a pitch, ask why not. The answer will tell you more than anything else in the meeting.
Entity thinking. GEO requires understanding your brand as a node in a knowledge graph: an entity with consistent attributes, relationships, and contexts that AI systems can retrieve and cite. This is a fundamentally different mental model from keyword targeting. Ask how they approach entity mapping for a new client. A team that has worked through this will have a specific, structured answer. A team that has not will reach for familiar SEO language and hope you do not notice.
The questions to ask before you hire a generative engine optimisation agency
The pitch meeting is your best diagnostic tool. These six questions produce responses that separate genuine capability from confident presentation.
How does my brand currently appear in ChatGPT and Perplexity?
A capable agency should be able to run this in the meeting. Open a laptop, type your brand name and two or three relevant prompts, show you the outputs, and explain what they see.
Good answer: they do it, they show you the raw AI response, and they frame the gap between current and target citation as the starting point for the conversation. Weak answer: "We include this in our onboarding process." If they cannot demonstrate the baseline before they have your money, you have no evidence they do this work at all.
What does your monthly reporting cover?
This question separates measurement from the appearance of measurement.
Good answer: branded mention rate across a defined prompt set, AI referral traffic segmented in GA4, comparison against the month-one baseline. Weak answer: organic traffic, keyword rankings, or a proprietary visibility index with no explanation of how it is calculated. A black-box score is not measurement. It is a number designed to prevent the question from being asked again next month.
Can you show me a client where AI citation share improved?
Push hard on this one. Agencies will reach for proxies if you allow it.
Good answer: the actual prompt and the AI's actual response, showing the client appearing in ChatGPT or Perplexity output before and after the engagement. Weak answer: a traffic graph, a vague reference to "significant citation improvement," or a case study where the metric is undefined. The specific prompt and the specific AI response is the evidence. Anything else is circumstantial. If they cannot produce this for a single client, ask whether they have any GEO clients yet. That is a legitimate answer. It is better than a graph dressed up as proof.
How does GEO fit with our existing SEO programme?
This question reveals whether the agency thinks in systems or in separate products.
Good answer: GEO builds on SEO infrastructure and the two run as a coordinated programme. If you have technical SEO gaps, those need addressing first or in parallel. Weak answer: GEO runs as a separate track independently of your organic SEO work. An agency that treats these as parallel silos will create gaps in your authority-building and duplication in your content work. They should not be separated.
What happens to results if we pause after six months?
This question probes whether the agency is building durable authority or running activity that requires the retainer to sustain it.
Good answer: the citation authority built through entity optimisation and digital PR tends to persist, in the same way domain authority in traditional SEO compounds over time. Ongoing monitoring prevents drift and competitor gains, but the structural work does not evaporate when the retainer pauses. Weak answer: "Results stop when the work stops." That suggests the programme is producing temporary effects rather than structural change. Know the difference before you sign.
Who on your team actually does this work and what is their background?
Agency pitches are often delivered by senior strategists. The work is often done by junior executives. Find out.
Good answer: specific names, specific roles, and relevant experience across technical SEO, structured data, content architecture, and digital PR. Weak answer: "Our specialist team" or "We have experts in every area." You are evaluating people, not a pitch deck. The person who implements your schema is not the same person who writes your media pitches. Ask whether both capabilities are genuinely in-house, or outsourced, and if the latter, to whom.
What GEO costs and when it makes sense
GEO in the UK is typically delivered as a monthly retainer. Credible programmes run between £2,000 and £8,000 per month, depending on scope, content volume, and how much digital PR activity is included. Programmes at the lower end tend to focus on content restructuring and entity work. Full-stack programmes with active digital PR outreach and dedicated citation monitoring sit at the higher end.
These figures come from public agency pricing guides and general market knowledge. If an agency quotes significantly outside this range in either direction, ask what is included or excluded. The outlier might be justified. It might also be a signal.
When GEO makes sense. You have meaningful domain authority and content that performs in organic search. You have genuine expertise and a clear brand identity worth citing. You want to extend your search visibility beyond traditional link results and into the AI answers your prospects are already reading. In these conditions, GEO adds an authority layer on top of foundations that are already sound. That combination compounds.
When GEO is not the right next step. If your site has thin content, weak domain authority, and no digital PR presence, the prerequisites for a GEO programme are not in place. You cannot build AI citation authority on a brand that AI systems cannot yet find credible third-party evidence for. The right move is to build the organic SEO and content foundations first, then layer GEO on top once there is something for AI models to reference. Any agency that tells you otherwise either does not understand the discipline or does not care about your outcome.
The ROI framing: AI referral traffic tends to arrive with prior intent already satisfied. The user has had their question answered by an AI and is visiting your site to act on that answer. The commercial dynamic is different from general organic traffic. The return on GEO investment builds slowly and then accelerates, similar to the compounding effect of organic SEO. The first three months are largely investment. The return arrives later and persists longer than campaign-style spend. Managing board expectations on that timeline is part of why starting with a clear baseline matters so much.
Where to start before you talk to any agency
The strongest position in any pitch meeting is knowing your current baseline before the agency presents theirs. This takes ten minutes and costs nothing.
Open ChatGPT, Perplexity, and Google AI Overviews. Type the prompts your prospects are actually using: who provides your service in your area, which companies are recommended for your specific offering, what should a business like mine look for in a provider. Does your brand appear? If not, where do your competitors appear? That gap is exactly what a GEO programme should address. Knowing it before the meeting means you can evaluate every proposal against a concrete brief, rather than accepting the agency's definition of the problem.
Check your foundations before you invest in the AI visibility layer. Do you have meaningful domain authority? Are you cited in relevant trade press or industry publications? Is your schema implemented correctly? A GEO programme built on weak SEO infrastructure will underperform regardless of how capable the agency is. Fixing the foundations first is not a delay. It is the right sequence.
Creative Tweed's AI Visibility Audit maps how your brand currently appears across the major AI engines, identifies the specific gaps, and shows you what is worth addressing first. It is the clearest way to walk into any agency conversation knowing what you need, rather than relying on someone with a commercial interest to define the problem for you.
GEO is not a standalone discipline. At Creative Tweed, it sits within a full search programme combining SEO, AI search, and conversion, so the authority built in one channel reinforces the others. If you want to understand what that looks like in practice, the AI SEO agency page explains the approach in detail.
The agencies worth hiring will welcome the fact that you have done your homework. The ones that do not are telling you something useful before you have spent a penny.
